This is a real implementation breakdown — clinic name anonymized at their request, but every number is real.
Westside Medical runs three locations in the Lower Mainland. At audit, they had:
- Two front-desk staff splitting 11 locations across a rotating schedule
- 40% of calls going to voicemail during peak hours (9–11am, 4–6pm)
- An average callback window of 6–18 hours
- A documented no-show rate of 28%
- Two staff spending roughly 60% of their time on phone/booking tasks
The brief was simple: reduce administrative load and no-shows without hiring.
What We Built
The system has three components:
1. AI Voice Agent
Built on a fine-tuned language model with the clinic's full booking logic, insurance acceptance matrix, and provider availability. Answers calls 24/7, books appointments directly into their EMR, and handles rebooking, cancellation, and FAQ.
Key customizations:
- Understands all three locations' provider schedules
- Handles the 14 most common insurance verification questions
- Escalates to human when it detects distress, complaint, or anything outside its defined scope
- Voice is warm, specific to their brand — not a generic robot
2. SMS Confirmation + Reminder Flow
48-hour and 2-hour reminders. Both include a one-tap reschedule link that goes back into the AI booking flow — not to a human. Cancellation triggers an automatic re-offer of the slot to a waitlist.
3. Missed Call Recovery
Any call that hits voicemail during office hours auto-triggers an SMS within 90 seconds: "Hi [name], we saw you called. Can we help? Reply with your question or a good time to reach you." Conversion from that message: 68%.
The Integration Work (What Actually Takes Time)
The tech gets demonstrated in 20 minutes. The integration takes 3 weeks.
Specifics for this clinic:
- EMR integration required custom webhook development (their system was legacy)
- Insurance verification required mapping their specific accepted plans to the AI's response logic — 94 insurance types
- The escalation logic required four iterations before it was right (early version escalated too aggressively; real staff were still getting interruptions for basic questions)
- PIPEDA compliance review added two days
Total timeline: 3 weeks from kickoff to live, 1 additional week of supervised operation before handoff.
What Broke (Honesty Section)
Two things failed in week one:
1. Multi-location slot conflicts. Our initial logic didn't account for a provider who works split shifts across two locations. The AI was double-booking. Fixed in 48 hours once caught, but this is why supervised rollout matters.
2. Voicemail detection false positives. The missed-call SMS was triggering for calls that were actually answered (some edge case in their phone system). Led to confused patients getting unsolicited texts. Resolved by changing the trigger to missed-call-with-no-subsequent-booking within 5 minutes.
Neither issue caused patient harm or significant operational disruption. Both were caught during the supervised week. This is not a reason to skip supervised rollout.
The Numbers at Month 3
Compared to pre-implementation baseline:
| Metric | Before | Month 3 |
|---|---|---|
| Call answer rate | 61% | 97% |
| No-show rate | 28% | 11% |
| Front desk time on phones | ~60% | ~18% |
| Avg callback window | 6–18 hrs | instant (AI) / 2 hrs (escalated) |
| Monthly admin cost savings | — | $4,200 |
The $4,200/month savings breaks down as:
- $2,800 in reduced front desk hours (staff reallocated to higher-value tasks)
- $900 in recovered revenue from bookings that would have been lost (voicemail drop-offs)
- $500 in no-show revenue recovery (re-filled slots from waitlist)
The system costs the clinic $380/month to run (API costs + maintenance retainer).
Net monthly ROI at month 3: approximately $3,820.
What the Staff Said
This is worth noting: the two front-desk staff were initially skeptical. Three months in, their feedback was positive — not because their jobs were threatened, but because the AI handles the repetitive, high-volume stuff (booking, reminders, FAQ) while they focus on the complex, in-person patient interactions that actually require human judgment.
One staff member's exact words: "I feel like I do actual work now."
What a System Like This Costs
For reference — a clinic-grade AI receptionist with EMR integration, custom insurance logic, and SMS flows runs:
- Setup: $8,000–$14,000 depending on EMR complexity
- Monthly: $250–$450 for API + infrastructure
- Optional maintenance retainer: $200–$400/month for ongoing optimization
Payback period at this clinic: 11 weeks.
Running a multi-location healthcare practice or service business with similar booking challenges? Book a Free Audit and we'll model the ROI for your specific situation before you commit to anything.